Metagenomics of pigmented and cholesterol gallstones: the putative role of bacteria

There is growing evidence for bacteria playing a role in the pathogenesis and formation of pigmented gallstones from humans. These studies mainly involved cultivation of gallstone-associated bacteria and 16S rRNA profiling, providing an indirect link between processes involved in gallstone formation by the bacteria in-situ. Here, we provide functional metagenomic evidence of a range of genes involved in bile stress response, biofilm formation, and anaerobic energy metabolism by Gram-negative Klebsiella in pigmented gallstones from a 76-year-old male patient. Klebsiella was also present in one cholesterol-type stone in a 30-year-old female patient who had additional cholesterol gallstones characterised by Gram-positive bacteria. Pigmented stones further revealed a predominance of genes involved in carbohydrate metabolism, whilst cholesterol stones indicated a profile dominanted by protein metabolism possibly reflecting known chemical differences between Gram-negative and Gram-positive biofilm matrices. Archaeal genes were not detected. Complementary carbon and hydrogen isotopic analyses of cholesterol within the patients’ stones revealed homogeneity, suggesting a common diet or cholesterol biosynthesis pathway that has little influence on microbial composition. This pilot study provides a framework to study microbial processes that play a potential role in gallstone formation across markedly different types of stones and patient backgrounds.

stones found in patient PM1 showed a high level of similarity at 80%. A 60% similarity in microbial diversity was observed between stones 1, 3 and 4 of patient CF4. Stone 2, however, only shared 40% similarity to the group (Fig. 1).
Metagenomic profiling of 4 replicate cholesterol gallstones of patient CF4 revealed a predominance of eukaryotes (96.05%) with the majority (98.5%) stemming from the patient or comprising human cells, with others (1.46%). Bacteria comprised only 3.6% of the ORFs. In agreement with the metagenomes of PM1 stones, a small portion of the ORFs were of viral origin (0.3%) and misidentified reads (0.05%), while no archaeal ORFs were recovered (Fig. 2).
As mentioned earlier, a consensus is yet to be drawn on what type of bacterial communities are common to pigmented and cholesterol gallstones 7 . This is in part due to the complex and unique interplay between a patient's health history, microbiome, environment, and a predominance of culture studies in which total community resolution is limited 7 . Nevertheless, in our study, patient PM1's pigmented stone bacterial community do approximately align with a previous culture analysis in which Klebsiella, Enterococcus, Enterobacter, E. coli and Psuedomonas aeruginosa were found to be the most prevalent genera across 61 pigmented and mixed gallstones 3 .
Less is known about microbial compositions in cholesterol gallstones, due to their propensity to have low amounts of bacterial DNA. However, this assertion may be due to the limitation of culture studies, as a recent investigation of the microbial composition of 27 cholesterol gallstones using high-throughput 16S rRNA profiling identified Enterobacteriaceae to be abundant 7 , whereas the less abundant genera differed from those present in patient CF4 in our study.

Functional metagenomic profiling. Genes involved in Bile Stress response. Microorganisms have been
known to survive and thrive in a range of extreme environments, including the human body where variations in pH, nutrient limitations, low oxygen levels, an established diverse bacterial microbiome, and host immunity responses provide longterm obstacles for survival [16][17][18] .
The human gallbladder environment poses its own unique challenges. The liver secretes approximately 800-1000 mL of bile into the gallbladder per day 19 . Bile acts as a detergent or emulsification agent for the digestion and absorption of lipids and contains sodium and potassium salts, bile acids (namely chenodeoxycholic acid, cholic acid), cholesterol, phospholipids, and bile pigments such as bilirubin 19 . This environment is made further hostile to bacteria as bile is concentrated 5-10 fold in the gallbladder, with commonly used antibiotic drugs, and heavy metals being secreted into bile as per the liver's detoxification and enterohepatic cycling processes 16,19 . The strategies bacteria employ to resist toxic agents such as bile and antibiotics are often complex and involve a variety of methods that include and are not limited to efflux pumps (which pump the toxic agent out of the cell), reduction of cell permeability, enzymatic modification or destruction of toxic agents either within or without the cell wall barrier, and the modification of the toxic agent's target either via genetic mutation, enzymatically, or by presenting an alternate target 20,21 . Bile, as a detergent, and consisting of a variety of toxic agents has been shown to cause membrane perturbations, DNA damage and oxidative stress in bacteria 21 . This is shown to be consistent with our current study where multidrug export efflux pumps, DNA and cell wall repair proteins were identified as important in bile resistance. Of the seventeen taxa in both patients combined that comprised at least 1% of the total sequenced gene pool, we identified six genera that harbour genes involved in bile stress survival, which are linked to the production of biofilms that may be associated with pigmented as well as cholesterol-type gallstone formation (Tables 2, 3 and 4) as discussed in detail below.
Gram-negative Bacteria. Klebsiella (bile-sensitive genes). The marABC and marR operons (for all genes identified see Table 2) are regulatory genes that control multiple antibiotic drug resistance 22 and have been shown to be activated in the presence of the bile salt deoxycholate, with the level of gene expression dependent on the salts' concentration 16,23 . The Tol protein and derivatives are important in many Gram-negative bacterial transport systems and act as an outer membrane pore function or efflux pump 24 . Mutations in Tol genes destabilise the membrane allowing for greater bile salt entry, thereby affecting bile resistance 24 . The TolC efflux pump (Table 2), in particular, has been shown to be upregulated in biofilms and is associated with the removal of toxic compounds and antibiotics 21 . The emrEB, mdtABCD, cmeAB genes similarly correspond to efflux pump action and  Figure 3. Heatmap with the major bacterial genera identified in the gallstones (n = 4) of patients PM1 (pigmented) and CF4 (cholesterol). The color key shows the relative abundance of the genera in the gallstones. The dendrograms illustrate the relationship between samples showing that the distribution of genera is relatively similar between replicate stones, but greatly differ between the two patients. are essential for bile resistance 24 . The emrEB multidrug efflux pump systems have been shown to actively efflux the bile salt chenodeoxycholic acid 25 . Furthermore, over-expression of mdtABCD, a multidrug resistance efflux pump cluster, leads to increased deoxycholate resistance 26 . The cmeAB has been shown to function as a multidrug efflux pump in C. Jejuni by effectively mediating resistance to bile salts 27 . The Dam (DNA adenine methylase) enzyme has been associated with repairing damage to Salmonella DNA after bile acid exposure 28 . Similarly, the sbcC and MutS (DNA repair), yvaG (rebuilding the cell membrane after stress), nifJ (oxidative response) and dgt (dGTP hydrolysis) homologs were associated with DNA and cell wall repair in response to bile stress in Enterococcus faecelis 29 .
Klebsiella (ORF's, promotors, proteins induced by bile). Experiments with bile salt treatments to Enterococcus faecelis 16,29 identified an increased production of a number of stress proteins (Gsp). Three of these stress proteins were identified in this study as the DnaK and GroEL/GroES molecular chaperones 29 and the organic hydroperoxide resistance protein Ohr 16,30 within Klebsiella (Table 3). Further experiments with Propinobacterium freudenraichii revealed genes involved in a variety of stress responses (heat, acid, bile salts) termed GSPs (General stress proteins), with bile salts in particular associated with oxidative stress responses 31 . A number of these were annotated to Klebsiella in this study and include the molecular chaperones Hsp20 (heat stress), DnaK, GoEL, AspA and ClpB that are associated with acid stresses 31 . SodA, an oxidative damage remediation gene, was also identified in this study and has been shown to be involved in stress responses within Lactobacillus lactis (oxygen stress), Bacillus subtilis (heat, salt and ethanol stresses), and B. cereus (heat, salt and ethanol stress) 31,32 . Further oxidative damage reduction and remediation proteins identified for Klebsiella were ORF002, G6PD, and the CysK, HemH homologs. The NADPH dependent aldo or keto-oxidoreductase ORF002 protein is an important part of the glutathione cellular defense system that is involved in the reduction of oxidative stress caused by reactive oxygen species (ROS) associated with bile 32,33 . The G6PD (Glucose-6-phosphate 1-dehydrogenase) protein has been shown to be activated in the presence of ROS, that arise due to stresses such as high levels of salt, and considered vital for cellular redox balance 34,35 . Similarly, the CysK (cysteine synthase) and HemH (ferrochetalse) homologs have also been shown to be overexpressed when exposed to bile-salt stresses 31,36 . Other acid stress proteins identified Genes disrupted in bile-sensitive mutants Function of gene products/putative function Reference(s)
were BCCP (a biotin containing carboxyl carrier protein) and ATPG (ATP synthase gamma chain) 31,37 . The DNA damage repair proteins MutB (Methylmalonyl-CoA mutase) and RecR (Recombinase) were also identified 16,36 . Genes encoding the dissimilatory dimethylsulfide reductase A,B and C (dmsABC, Table 3) were retrieved in Klebsiella, indicating its capacity for anaerobic metabolism and use of dimethylsulfoxide (DMSO) as a terminal electron acceptor. DMSO respiration is energetically favourable under anaerobic conditions in bacteria that contain this metabolic potential 38 . Further, the dmsABC operons are controlled by the oxidative regulator fnr, important for oxidative stress response and anaerobic metabolism in pathogenic bacteria [39][40][41] . This ability by Klebsiella may explain its successful survival and growth in the anoxic conditions present in the human gallbladder and its dominance in the present study. Other gram-negative and gram-postive bacteria identified in this study may be out competed by Klebsiella or utilise fermentation for energy metabolism instead, a less efficient form of energy conservation than DMSO 42 .
Escherichia (bile-sensitive genes). Similar to Klebsiella, Escherichia exhibited the oxido reductase gene yvaG or membrane composition and repair protein in our study (Table 2).

ORFs/promotors/proteins induced by bile
Function of gene products/putative function Reference(s)
Shigella (bile-sensitive genes). Shigella exhibited the PhoQ regulatory protein ( Table 2). PhoQ is closely associated with the PhoP regulon. The combined PhoP-PhoQ proteins have been associated with various bacteria and their ability to sense and resist bile stress 43 . Bacterial mutants missing PhoP-PhoQ were killed at significantly lower concentrations of bile than those with these proteins, and those with PhoP alone surving a >60% concentration of bile in lab conditions 43. Shigella (ORF's, promotors, proteins induced by bile). Similar to Klebsiella and Escherichia, Shigella exhibited the NADPH dependent aldo or keto-oxidoreductase ORF001 and ORF002 proteins involved in bile oxidative stress reduction (Table 3).  Genes involved in biofilm production. In addition to mediating toxic substances via the resistance strategies described above, microorganisms can group together, attach to either living or non-living surfaces, and form what is called a biofilm 45 . Biofilms comprise a variety of microorganisms enclosed in an extracellular polymeric substance (EPS) or matrix made up of mostly polysaccharides and other environmental specific materials 45 . Biofilm formation has considerable advantages and has shown to protect bacterial communities from UV light, heavy metals, acidity, hydration or salinity changes and host immune responses, including large doses of antimicrobial drugs that would be lethal to the same community if in a planktonic state 21,46 . There are three main stages involved in biofilm formation: initial adherence to a surface, development of a community structure and ecosystem, and eventual detachment 45 . Each stage is a complex process regulated by a variety of genes that are often environmental, bacterial strain or stressor specific 21 . Common mechanisms include the development of curli fimbriae (adherence or attachment mechanisms), and quorum-sensing or cross community communication to coordinate biofilm attachment, development, detachment, and resistance 13,46 . The main resistance mechanisms afforded by the EPS matrix include drug indifference, in which the EPS works as a barrier between the drug and the targeted microbial cell membrane, the allowance of antibiotics to slowly diffused through the EPS, so that time is given for resistance mutations to develop, efflux pumps, and the secretion of periplasmic glutans that keep toxic substances away from intracellular targets 21 . Certain organisms may provide the base biofilm, whilst others live either competitively or symbiotically within it, with environmental and community composition in a state of constant change 46 . As previously reported, biofilms with Salmonella were identified through SEM and culture analysis, on the surface of cholesterol gallstones 6 . Our study also suggested the presence of a biofilm on the surface of both pigmented and cholesterol type gallstones since stringent UV sterilisation initially resulted in a reduction of the yield of extracted genomic DNA by over 90%. Indeed the presence of a microbial biofilm on the surface of the patients' gallstones was confirmed by taxonomic and functional metagenomics analysis. The following genera were identified in the gallstone metagenomes that harbor genes which are putatively associated to biofilm production.

Serratia (ORF's, promotors, proteins induced by bile
Klebsiella. The CsgD gene is a transcriptional activator involved in the regulation of curli fimbriae biosynthesis 13 (Table 4). Curli fimbriae have been identified to be significant EPS components within Enterobacteriaceae and involved in bacterial adherence to abiotic substances and cell adhesion during symbiotic or infectious processes 13 .
Klebsiella, an enterobacterium, has exhibited the CsgD gene, alongside the Type 1 and Type IV fimbriae in this study. The Type 1 and Type IV fimbriae are part of the gene cluster fim, containing all the components required for fimbrial assembly, and associated with capsule and pilin processes -significant factors in colonisation and biofilm production 17 . Type 1 and Type IV fimbriae have also been shown to facilitate biofilm assembly on both abiotic and host-derived extracellular matrix protein surfaces 13 . The wza and wzc genes encode for surface molecules involved in capsule assembly and are considered to be important in the early stages of biofilm formation by Klebsiella pneumoniae 13,17 . A study isolating the genes involved in biofilm formation of the K. pneumoniae strain causing Pyogenic Liver Abscess found SugE an important gene that affects biofilm production by modulating capsular polysaccharide production and biofilm mucoviscosity 17 . The ClpX and LuxR regulatory genes, part of the sugar-specific phosphotransferase systems, and the cold shock protein cspD were also implicated in biofilm production 17 .
Recently, various strains of Klebsiella were tested to determine those with the highest biofilm production and the genes associated with this process 13 . The strain identified with the highest output of biofilm implicated the RbsA and RbsC genes alongside the quorum sensing molecule Autoinducer 2 (AI-2) and the prophage CP4-57 integrase as putitavively involved in the process 13 . These important genes associated with Klebsiella were also identified in our study (Table 4) The RapA gene identified has been shown to play a role in regulating the yhcQ gene that encodes a putative multidrug efflux pump and yeeZ, a gene associated with biofilm production 21 . Polyphosphate kinase has been linked with biofilm development, quorum sensing and virulence in P. aeruginosa and was also annotated to Klebsiella in our study 47,48 .
Escherichia. The surface molecule encoding wzb and wzc genes, important for capsule assembly and early stage biofilm formation in K. pneumoniae 13,16 , were also associated with Escherichia in our study (Table 4).
Enterococcus. Similar to genes involved in biofilm formation (the RbsA, SugE, ClpX, LuxR, CspD) discussed for Klebsiella above were associated with Enteroccoccus in our study. Enterococcus also exhibited the galE gene shown to influence lipopolysaccharide structure, colonisation and biofilm formation 49 .
Other relevant cellular processes. Consistent with a more prominent presence of number of species and genes encoding for processes associated with resistance to oxidative stress from bile and biofilm production, patient PM1 also exhibitied an overall higher abundance of genes involved in stress response, cell wall and capsule production, cluster-based subsystem activity as well as carbohydrate metabolism compared to patient CF4 (Fig. 4). A potentially enhanced carbohydrate metabolism in PM1 may be attributed to low levels of nutrients in the gallbladder resulting in microbes to metabolise excess biofilm (namely polysaccharides), as reported previously 50 . In contrast, Gram-positive bacteria, mainly associated with gallstones of patient CF4, reveal a higher relative abundance of genes involved in protein metabolism. One hyothesis is that these bacteria are involved in the decomposition of dead human cells associated with the gallstones of this patient (Fig. 4) as inferred from the high relative abundance of human genes compared to PM1. Another explanation for the observed diffence in a carbohydrate vs. protein dominated microbial metabolism between the gallstone types is that biofilms of Gram-negative and Gram-positive bacteria differ chemically from each other. EPS produced by Gram-negative bacteria, which predominate in PM1, exhibit anionic properties (attributed to uronic acids) that enable calcium and magnesium ions to bind with polymer strands providing a more tightly bound biofilm architecture 45 . Gram-positive bacteria (mainly in CF4) have been shown to exhibit a more cationic EPS charge, and be composed of teichoic acid mixed with small quantities of protein 45,51 , which may explain a higher relative abundance of genes involved in protein metabolism.
Cholesterol analysis. We conducted compound specific isotopic analysis (CSIA) of the patients carbon (δ 13 C) and hydrogen (δD) isotopes of cholestrol to ascertain possible dietary or exogenous environmental factors that may be associated with or divergent from the bacteria identified in this study. Individual compounds in a complex mix (i.e. a gallstone made from patient-specific cholesterol/bile mixtures) can have distinct differences in their carbon, hydrogen, oxygen, nitrogen or sulfur isotopic signatures 52 . The marked difference between the isotopic weight of δ 13 C of an identified compound in two gallstones, for example, can indicate that a different source or mechanism was utilised during the synthesis of the compound 52 . We investigated whether or not the patients' native stones were homogenous and if homogeneity existed between the patients themselves.
The only molecular component identified by GC-MS within the stones of each patient was the Cholest-5-en-3β-ol compound with trace amounts of 5α-Cholest-7-en-3β-ol. Subsequently, CSIA was conducted on the non-derivatised Cholest-5-en-3β-ol compound of each of the patients' stones ( Table 5).
The Cholest-5-en-3β-ol δ 13 C values obtained for patient PM1 showed a range between −24.7 and −25.9‰, and a range between −23.1 and −23.7‰ for patient CF4, resulting in an approximately 1‰ difference across the four stones analysed for each patient. The Cholest-5-en-3β-ol δD showed a value range of −218 and −231‰ for patient PM1, resulting in an approximately 13‰ difference across the four stones analysed for this patient. The Cholest-5-en-3β-ol δD showed a value range of −252 and −254‰ for patient CF4, resulting in a negligible 2‰ difference across the four stones analysed for this patient.
Within each patient both the δ 13 C and δD values for Cholest-5-en-3β-ol were not significantly different amongst the 4 stones analysed, supporting a common source for Cholest-5-en-3β-ol. Between the two patients the δ 13 C and δ D values differed only by minor amounts also supporting a common source for Cholest-5-en-3β-ol.

Conclusions
This pilot study explored taxonomic and functional metagenomics and sterol homogeneity within two patients of diverse backgrounds to elucidate a possible universal factor at play in gallstone pathogenesis and formation. For the first time, functional genes were identified that were associated with bile stress response and biofilm development as possible microbial processes leading to the formation of both pigmented and cholesterol-type gallstones. In the analysed pigmented stones, genes involved in biofilm formation were mainly recovered from clinically pathenogenic Klebsiella and Enterococcus while bile resistance genes were present also in Escherichia, Shigella, Serratia and Bacillus. Klebsiella was also present in one of the chlolesterol gallstones, while the remaining analysed cholesterol stones showed a predominance of Gram-positive bacteria that were not identified within the pigmented stones. Klebsiella was also the only genus to exhibit DMSO respiration, giving it a distinct advantage in the anoxic environment of the human gallbladder. This, in conjuction with being the genus to exhibit the highest number of genes involved in bile stress response, and biofilm formation, may place Klebsiella as a major player in gallstone pathanogenesis. Further, pigmented stones, predominated by Gram-negative bacteria, revealed a high proportion of genes involved in carbohydrate metabolism, whilst cholesterol stones indicated a profile dominanted by protein metabolism. A possible explanation for the observed diffence in a carbohydrate vs. protein dominated microbial metabolism between the gallstone types is that biofilms of Gram-negative and Gram-positive bacteria differ chemically from eachother resulting with the latter having a higher protein content in the EPS matrix. Fungal and archaeal genes were not detected in both types of stones. Complementary carbon and hydrogen isotopic analyses of cholesterol within the patients' stones revealed homogeneity, suggesting a common diet or cholesterol synthesis pathway that has only a minor influence on microbial composition.
This pilot study provides a framework to study microbial processes that play a potential role in gallstone formation across markedly different types of stones and patient backgrounds. In addition, future studies could also involve metatranscriptomic profiling to ultimately reveal which bacteria are actively expressing genes involved in processes such as bile stress response and biofilm formation that could contribute to the pathenogensis of gallstones.

Methods
Sample collection. Samples were collected whilst the patients were undergoing laproscopic cholecystectomy and were immediately rinsed in sterile saline solution (9 g L −1 NaCl) and placed in sterile glass containers. The samples were immediately stored at −80 °C until further processing. All patients provided written informed consent upon enrolment to the study. Quantitative PCR. To quantify the amount of bacterial 16S rRNA gene copies, an aliquot of the extracted and purified genomic DNA was subjected to quantitative polymerase chain reaction (qPCR) using general primers 55 targeting the V4 region of bacterial 16S rRNA. All reactions were performed using SYBR Premix Ex Taq (TLi RNase H Plus) (Takara Bio Inc) in a Realplex quantitative PCR cycler (Eppendorf) and involved initial denaturing (1 min at 95 °C), followed by 32 cycles including denaturing (5 s at 95 °C), primer annealing (30 sec at 60 °C), primer extension and imaging of newly formed fluorescent (SYBR ® green I labelled) double-stranded DNA (72 °C for 60 sec). Between 10 1 and 10 6 copies (10-fold dilution series) of bacterial 16S rRNA were added to reaction mixtures and served as standards during qPCR to calibrate the copy numbers of bacterial 16S rRNA in the gallstone samples.

Processing of sequence data and bioinformatics.
Approximately 280 million paired-end sequence reads (see Supplementary Table S1) were imported into CLC Genomics Workbench 8.0 (CLC Bio) and trimmed of ambiguous reads to a quality limit of 0.5. Contigs were assembled using the CLC Genomics Workbench paired-end Illumina (de novo) read assembler with automatic bubble and word size, length fraction of 0.5, similarity fraction of 0.95, and a minimum contig size cut-off of 300 nucleotides. Contigs were assembled without scaffolding to reduce the formation of chimaeric assemblies. The CLC Genomics Workbench read mapping option was used to map reads onto contigs. ORFs within the contigs were detected using FragGeneScan 8 . Taxonomic assignments of contigs were performed using the NCBI BLASTp software suite against the SEED database of predicted proteins from cultivated microbial genomes with assigned taxonomy. The basis for taxonomic assignment of the ORFs was amino acid similarity of >60% over an alignment length of >50 amino acids to predicted proteins present in the database with an assigned taxonomy 8 . A matrix showing the relative abundance (average coverage) of annotated ORFs deriving from specific taxa per sample was produced using a python script publicly available online (bitbucket.org/wrf), and was subsequently used for downstream analysis. Heatmaps were performed in R (http://www.r-project.org/) using the vegan (http://vegan.r-forge.r-project.org/) and the Bioconductor Heatplus (https://bioconductor.org/biocLite.R) package. The data was normalized, with the Hellinger function used to produce the taxonomy overview heatmap to show species that may have been obscured by the dominant reads (Fig. 3). The overview of Subsystems, Level 1, cellular processing category annotations were obtained from the SEED database via MG-RAST (Project ID: mgp81110-81111; metagenomics.anl.gov). The Primer-E software package (http://www.primer-e.com/) was used to generate principle coordinate analysis (PcoA) plots using the Bray-Curtis distance metric.
Gas Chromatography Isotope Ratio Mass Spectrometry. Four gallstones from each patient were individually crushed in heat-sterilized (500 °C, 8 h) mortars. The grounded powder was then extracted via sonication (1 h) with dichloromethane (DCM) and methanol (9:1). The extracts were then fractionated by small-scale column liquid chromatography 56 . Approximately 2 mg of the total extract was placed on top of a small column (5 × 0.5 cm i.d.) of activated silica gel (160 °C, 8 h). The first hydrocarbon fraction was eluted with n-hexane (2 mL), the second hydrocarbon fraction with DCM in n-hexane (1:4, 2 mL), and the more polar fraction with an equal mixture of DCM and methanol (1:1, 2 mL). The fractions were analysed by gas chromatography-mass spectrometry (GC-MS).
The polar fractions (containing Cholest-5-en-3β-ol) were each analysed by compound specific isotope analyses to obtain δ 13 C and δD values of Cholest-5-en-3β-ol. The instrument used was a Thermo Delta V Advantage isotope ratio monitoring mass spectrometer (irMS), coupled to a Thermo Trace GC Ultra via a GC Isolink and Conflo IV. The column used was an Agilent DB-5MS Ultra-Inert, 60 m long, 0.25 mm (i.d.), with 0.25 µm film thickness. An aliquot of 1 µL of each fraction was injected into the split/splitless injector in splitless mode, held at 280 °C. The GC oven was increased from 40 to 325 °C at 10 °C/min, then held at 325 °C for 10 min. The carrier gas used was helium held at a constant flow of 1.5 mL/min.
For the carbon isotope analysis, GC column outflow was passed through the GC Isolink combustion reactor (copper oxide / nickel oxide, 1000 °C) to combust hydrocarbons to CO 2 . For hydrogen isotope analysis, the outflow passed through the high-temperature conversion reactor (graphite-lined, 1420 °C) and was pyrolysed to H 2 . The CO 2 / H 2 passed through the Conflo IV interface to the irMS, which measured m/z 44, 45 and 46 (for CO 2 ) or m/z 2 and 3 (for H 2 ). The δ 13 C and δD values were calculated from the measured masses by Thermo Isodat software, and calibrated to the VPDB (for CO 2 ) and VSMOW (for H 2 ) scales by comparison with a mixture of n-alkane standards of known isotopic composition.
Data availability. Data